Economics 50: Intermediate Microeconomics
Summer 2010
Stanford University
Michael Bailey
Lecture 3: Deriving Demand, Consumption Curves, and WARP
Overview
The optimal bundle is at the point where the indierence curve is tangent to the budget line
The optimality condition breaks down when preferences are not monotone or strictly convex
Even when preferences are strictly convex and monotone, the Lagrangian Method needs to be amended
for corner solutions
Demand depends on the shape of the budget line
Properties of preferences and demand alter how demand changes with income and prices
Revealed preferences allows us to derive preferences from optimal choices
Demand
To nd the optimal bundle demanded by the consumer, we need to know the preferences over the bundles
in the choice set and the budget constraint. The consumer will choose the most preferred bundle that is in
the budget set. If preferences are complete and transitive, and can be represented by a utility function u(x),
then we can equivalently state the consumer problem as "nd the bundle that maximizes utility subject to
the budget constraint", written mathematically:
x
= arg max u(x
1
; x
2
; :::) s:t:
1
i=1
p
i
x
i
_ I and x
i
_ 0\x
i
To solve this problem, it is often convenient to use the Lagrangian Method. The lagrangian for this
problem would be (ignoring the non-negativity constraint on the goods):
/(x
1
; x
2
; :::; ) = u(x
1
; x
2
; :::)
_
1
i=1
p
i
x
i
I
_
1
which has rst order conditions
1
:
@/(x
1
; x
2
; :::; )
@x
i
=
@u(x
1
; x
2
; :::; )
@x
i
p
i
= MU
xi
p
i
= 0 \i
@/(x
1
; x
2
; :::; )
@
=
_
1
i=1
p
i
x
i
I
_
= 0
Solving for :
=
MU
xi
p
i
\i
Recall that the lagrange multiplier is the marginal benet of relaxing the constraint by one unit, or in
this consumer problem, giving the consumer one more dollar of income. How much additional benet would
the consumer get out of an additional dollar? If it spent on buying more of good x
i
then the additional benet
would be the utility gained times by the amount of x
i
that could be bought for one dollar, MU
xi
(
1
pi
) =
MUx
i
pi
. At the optimal bundle, an additional dollar of income has marginal value of
MUx
i
pi
to the consumer.
An even more important result in terms of the intuition of the consumer problem is found by rearranging
the rst order conditions:
=
MU
xi
p
i
=
MU
xj
p
j
==
MU
xi
MU
xj
=
p
i
p
j
== MRS =
p
i
p
j
== MRS =
p
i
p
j
== Slope of indierence curve = Slope of budget line
What this result tells us is that at the optimum, the indierence curve will be tangent to the budget line.
Suppose this wasnt the case and at the optimum the indierence curve crossed the budget line. Because
indierence curves are level curves (think contour lines on a topographical map), we could move up or down
the budget line and gain in utility as one side of the indierence curve will have higher utility (the consumer
could "climb" to a higher utility). Thus the optimal bundle must not have been optimal to begin with. The
only way the consumer would not be able to "climb" to a higher indierence curve is if it was tangent to the
budget line, meaning the direction of higher utility would be outside of the budget set.
1
These conditions are necessary for nding a maximum but are not sucient. In this class, we will not worry about the
suciency conditions for the Lagrangian Method.
2
Figure 1: The bundle located where an indierence curve crosses the budget line cannot maximize utility
Figure 2: The bundle that maximizes utility is where the budget line and the indierence curve are tangent
3
Example 1 Suppose that utility is given by the Cobb-Douglas form u(x; y) = x
a
y
b
and prices and income
are given by p
x
, p
y
; and I respectively. What is the individual demand for x and y denoted x
(p
x
; p
y
; I)
and y
(p
x
; p
y
; I) that maximize utility subject to the budget constraint? Graph the optimal bundle on the
indierence curve and on the budget line.
max x
a
y
b
s:t: p
x
x +p
y
y = I
/ = x
a
y
b
(p
x
x +p
y
y I)
==
ax
a1
y
b
bx
a
y
b1
=
a
b
y
x
=
p
x
p
y
== y = x
p
x
p
y
a
b
== p
x
x +p
y
x
p
x
p
y
a
b
I = 0 subbing into BC
== p
x
x +p
x
x
a
b
I = 0
== p
x
x +p
x
x
a
b
= I
== p
x
x(1 +
a
b
) = I
== p
x
x(
a +b
a
) = I
== p
x
x =
a
a +b
I
== x
(p
x
; p
y
; I) =
a
a +b
I
p
x
== y
(p
x
; p
y
; I) =
b
a +b
I
p
y
by symmetry of the problem
Notice that the demand functions for C-B imply that the consumer will spend a constant share of income
on x and y: p
x
x
=
a
a+b
I and p
y
y
=
b
a+b
I: Note that x
and y
are normal goods.
4
0 1 2 3 4 5 6 7 8 9 10
0
1
2
3
4
5
6
7
8
9
10
x
y
The optimal bundle (x
; y
) = (
5
4
;
5
2
) occurs where the budget line is tangent to the indierence curve
When Does the Lagrangian Method Fail to Find the Maximum?
The lagrangian solution is nice and elegant and the economic intuition of the solution and multiplier is an
economists dream. Sadly, there are many cases where the lagrangian could possibly fail to identify the
utility-maximizing bundle, or even identify a bundle that is not a maximum. These cases, which we shall
investigate separately, are:
Non-monotone preferences
Non-dierentiable points on the indierence curve
Concave preferences
Weakly convex preferences
Corner solutions
Non-monotone preferences
Suppose that utility is not increasing in one of the goods. The bundle of highest utility would have none of
this good included, but the lagrangian method might identify a bundle with some of that good included. For
example, consider the preferences given by u(x; y) = x
a
y
b
(a decreasing monotonic transformation): Notice
that the indierence curves would be identical to those of Cobb-Douglas preferences, but utility would be
increasing towards the origin, instead of in the standard monotonic direction. The optimal bundle would
5
be at the origin x
; y
= (0; 0), but the Lagrangian Method would nd the same solution as before. Also,
remember that the requirement for the consumer to consume their entire income p
x
x + p
y
y = I is that
preferences are monotone.
If utility is always decreasing in one good, one can just drop that good from the problem (set x
i
= 0) and
resolve. For example, if u(x; y) = y x; then MU
x
< 0 for all values of x; so we set x
= 0 and maximize
the utility function u(x; y) = y: If utility is decreasing over a range of values for some good, then the optimal
amount of that good could not be within that range, otherwise utility could be increased by decreasing the
amount of that good consumed.
Figure 3: The lagrangian bundle does not maximize utility because preferences are not monotone, the optimal
bundle is at (0; 0)
Example 2 u(x; y) = xy 10x
Notice that MU
x
= y 10 is positive whenever y > 10; thus the optimal bundle x
; y
would have x
= 0
whenever y
> 10:
Non-dierentiable points on the indierence curve
Suppose preferences are given by min(ax; by) =
(ax+by)jaxbyj
2
which is not dierentiable at ax = by and
indierence curves are "kinked" at that point. Since the kink does not have a slope, using the lagrangian to
nd the point where the slope of the IC is the same as the slope of the BL is not going to work.
Note that for perfect complements, the optimal bundle is always going to occur at ax
= by
; i.e. the
ratio of x to y will be constant:
a
b
=
y
. This is because if the consumer is consuming the bundle x; y such
6
that ax = by, then additional x will bring no additional utility, and additional y will bring no additional
utility, as min(ax; by) will be unchanged. We can thus solve for the optimal demand functions by subbing
ax
= by
into the BC :
p
x
x +p
y
_
a
b
x
_
= I
== x
_
p
x
+p
y
a
b
_
= I
== x
(p
x
; p
y
; I) =
I
p
x
+p
y
a
b
== y
(p
x
; p
y
; I) =
I
b
a
p
x
+p
y
Graphically, notice that the highest indierence curve we can reach is where the budget line passes
through the "kink" in the indierence curve, where ax = by:
Figure 4: Demand for perfect complements
Concave preferences
Consider the utility function given by u(x; y) = x
3
+y
3
: The preferences are strictly monotone, but not convex
_
@MRSx;y
@x
=
2
xy
2
> 0
_
: The Lagrangian Method would indicate that x
; y
=
_
I
py
q
py
px
+px
;
I
px
q
px
py
+py
_
: If
p
x
= p
y
= 1; then the lagrangian bundle would be given by
_
I
2
;
I
2
_
; or spend half of your income on each
of the two goods. Notice that u
_
I
2
;
I
2
_
= 2
_
I
2
_
3
=
I
3
4
: Is this the optimal bundle? Consider the deviation
where all income was spent on x : x; y =
_
I
px
; 0
_
which has utility u(I; 0) = I
3
at p
x
= p
y
= 1: This yields
four times as much utility as the lagrangian bundle (I
3
>
I
3
4
): Why did the lagrangian fail? Note that the
7
lagrangian method did indeed nd the bundle where the IC was tangent to the BL; but an even higher
indierence curve could be reached by consuming a bundle on the endpoints.
When preferences are concave, rather than "averages are preferred to extremes", we have "extremes are
preferred to averages", so the lagrangian method which nds the point where the marginal benet from
consuming each good is equal will nd an average of sorts when utility would be higher by consuming only
one of the goods. This is often the case when the marginal utilities are increasing; it pays to "put all your
eggs in one basket".
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8 3.0
0
1
2
3
x
y
Figure 5: u(x; y) = x
3
+ y
3
, p
x
= p
y
= 1; and I = 2: Note that utility is not maximized at the tangency
point for this concave function, the endpoint bundles lie on a higher indierence curve.
Weakly convex preferences
If indierence curves are straight lines, then there will either be no points, or innitely many points where
the slope of the indierence curve equals the slope of the budget line. Usually, the optimal bundle will occur
where the consumer buys all of one good. If indierence curves have a linear segment, either the optimal
bundle will occur at a tangency where the indierence curve is non-linear, or at an endpoint of the linear
segment.
Example 3 Perfect Substitutes u(x; y) = ax +by
If we attempt to use the Lagrangian Method, we end up with a rst order condition:
a
b
=
px
py
; which
consists entirely of exogenous variables, so does not tell us anything about the optimal solution. If we graph
a set of indierence curves, we see that if the ICs are steeper than the BL
_
px
py
>
a
b
_
, then the highest
indierence curve that intersects the BL is where y
= 0: Similarly, if the BL is atter than the ICs
8
_
px
py
<
a
b
_
, then the highest IC intersecting the BL occurs where x
= 0: If the BL and ICs have the
same slope, then the BL will coincide with an IC and the consumer is indierent between all bundles on the
BL: Hence, our optimal demand functions are given by:
x
(p
x
; p
y
; I); y
(p
x
; p
y
; I) =
(
I
px
; 0) if
px
py
<
a
b
(0;
I
py
) if
px
py
>
a
b
(z
I
px
; (1 z)
I
py
) for some z [0; 1] if
px
py
=
a
b
Figure 6: Demand for perfect substitutes when
px
py
<
a
b
Remark 4 Every time you derive a demand function, check whether the demand is elastic/inelastic, the
goods are normal, and the goods are complements/substitutes.
9
Figure 7: Demand for perfect substitutes when
px
py
<
a
b
Figure 8: Demand for perfect substitutes when
px
py
=
a
b
: The consumer is indierent between all bundles on
the budget line
10
Corner solutions
So far, it looks like there are no problems when we are nding the maximum of a strictly convex function when
preferences are strictly convex. Consider the quasi-linear function u(x; y) = 4
_
x+y: Verify that preferences
are strictly monotone and convex. To solve for the demand, use the standard rst-order condition:
2x
1=2
1
=
p
x
p
y
== x
(p
x
; p
y
; I) =
_
2
p
y
p
x
_
2
== p
x
_
2
p
y
p
x
_
2
+p
y
y = I
== y
(p
x
; p
y
; I) =
I 4
p
2
y
px
p
y
=
I
p
y
4
p
y
p
x
However, y
(p
x
; p
y
; I) is positive only when I > 4
p
2
y
px
: This is because the tangency point is now occurring
below the x-axis.
1 2 3 4 5 6 7 8 9 10
-5
-4
-3
-2
-1
0
1
2
3
4
5
x
y
u(x; y) = 4
_
x y, p
x
= p
y
= 1; and I = 2: The lagrangian demand is x
; y
= (4; 2) where the IC is
tangent to the BL
The optimal bundle is not attainable as consumption is constrained to be positive. As we did not
incorporate the non-negativity constraints into the lagrangian rst order condition, the optimality condition
is not constraining consumption to be positive. Notice that in this problem, the highest indierence curve we
could attain given the budget constraint is by spending all income on x : x
; y
= (
I
px
; 0): When all income
is spent on one good, this is known as a corner solution. The lagrangian works ne with strict convexity
and strict monotonicity as long as the solution is an interior solution. Incorporating the non-negativity
11
constraint into the lagrangian is the subject of the next section.
Incorporating Non-negativity Constraints
The rst order conditions for the lagrangian do not account for the non-negativity constraint x
i
_ 0\x
i
: We
must amend the rst order conditions in the following way:
@/(x
1
; x
2
; :::; )
@x
i
=
@u(x
1
; x
2
; :::; )
@x
i
p
i
= MU
xi
p
i
_ 0 \i
x
i
(MU
xi
p
i
) = 0 \i
@/(x
1
; x
2
; :::; )
@
=
_
1
i=1
p
i
x
i
I
_
= 0
Note that if the i
th
rst order condition does not hold with equality, we set x
i
= 0: These are known as
the Kuhn-Tucker conditions.
Example 5 Quasi-linear Utility: u(x; y) = 4
_
x +y has the following rst-order conditions:
@/(x; y; )
@x
=
@u(x; y)
@x
p
x
= 2x
1=2
p
x
_ 0
@/(x; y; )
@y
=
@u(x; y)
@y
p
y
= 1 p
y
_ 0
@/(x; y; )
@
= p
x
x +p
y
y I = 0
If we assume an interior solution such that the FOCs hold with equality, we derive the demands x
(p
x
; p
y
; I) =
_
2
py
px
_
2
and y
(p
x
; p
y
; I) =
I4
p
2
y
px
py
=
I
py
4
py
px
: We found that when (p
x
; p
y
; I) = (1; 1; 2); then x; y = (4; 2)
satises the rst order conditions.
x; y = (2; 0) also satises the FOC: Notice that when y = 0; x =
I
px
to satisfy the budget constraint.
Because x > 0; we need the rst rst-order condition to hold with equality, otherwise the condition x (MU
x
12
p
x
) = 0 would fail to hold :
2x
1=2
p
x
= 0
=
2x
1=2
p
x
=
2
_
I
px
_
1=2
p
x
=
2
_
_
I
px
_
1
p
x
= MU
x
1
p
x
This makes sense, if y = 0 then the lagrange multiplier should only reect the value of an additional unit
of income in terms of x: Finally, we check that the second rst order condition is not positive 1 p
y
=
1
2
q
(
I
px
)
py
px
_ 0: This will hold as long as
1 _
2
_
_
I
px
_
p
y
p
x
== 1 _
4
_
I
px
_
_
p
y
p
x
_
2
== I _
4p
2
y
p
x
which is our original condition for when y
< 0:
Playbook for Deriving Demand
1. Check if preferences are monotone. If utility is decreasing in one of the goods x
i
; then set x
i
= 0 and
resolve the problem.
2. Check if preferences are strictly convex. If preferences are concave or weakly convex, then the optimal
bundle is probably at a corner solution.
3. If preferences are strictly monotone and convex, use Lagrangian Method. If you nd that the optimal
bundle occurs where x
i
< 0; then set x
i
= 0 and nd a new optimum using Kuhn-Tucker conditions.
In the two good case, it is sucient to set x
i
= 0 and x
j
=
I
pj
:
Example 6 max u(x; y) = xy 10x s:t: p
x
x +p
y
y = I
Notice that MU
x
= y 10 is positive whenever y > 10; thus the optimal bundle x
; y
would have x
= 0
13
whenever y
< 10: MRS =
y10
x
so this function is strictly convex whenever y > 10: We proceed tentatively
with the Lagrangian method noting that if y < 10; then the consumer will spend all income on y:
MRS =
y 10
x
=
p
x
p
y
== y =
p
x
p
y
x + 10
== p
x
x +p
y
_
p
x
p
y
x + 10
_
= I Subbing into BC
== x
(p
x
; p
y
; I) =
I 10p
y
2p
x
y =
p
x
p
y
_
I 10p
y
2p
x
_
+ 10
== y
(p
x
; p
y
; I) =
I 10p
y
2p
y
+ 10
Note that x
> 0 ==
_
I > 10p
y
which also guarantees that y > 10: We thus know that optimal demand
will be given by:
x
(p
x
; p
y
; I); y
(p
x
; p
y
; I) =
_
I10py
2px
;
I10py
2py
+ 10
_
if
_
I > 10p
y
(0;
I
py
) if
_
I _ 10p
y
1 2 3 4 5 6 7 8 9 10
-20
-10
0
10
20
30
40
50
x
y
Figure 9: Set of indierence curves for u(x; y) = xy 10x
Example 7 max u(x; y) = 100 lnx +y s:t: p
x
x +p
y
y = I
14
Utility is strictly convex and strictly monotone.
MRS =
100
x
=
p
x
p
y
== x
(p
x
; p
y
; I) = 100
p
y
p
x
== 100p
y
+yp
y
= I Subbing into BC
== y
(p
x
; p
y
; I) =
I 100p
y
p
y
x
is always positive, y
is positive whenever I > 100p
y
: Demand for x and y will be given by:
x
(p
x
; p
y
; I); y
(p
x
; p
y
; I) =
_
100
py
px
;
I100py
py
_
if
_
I > 100p
y
(
I
px
; 0) if
_
I _ 100p
y
0 20 40 60 80 100 120 140 160 180 200 220 240
0
50
100
150
200
250
x
y
Figure 10: u(x; y) = 100 ln(x) +y; (p; I) = (2; 1; 150) : Interior solution x
; y
= (50; 50)
15
0 20 40 60 80 100 120 140 160 180 200 220 240
0
50
100
150
200
250
x
y
Figure 11: u(x; y) = 100 ln(x) +y; (p; I) = (2; 1; 100) : Corner solution x
; y
= (50; 0)
Funky Budget Constraints
So far we have only considered maximizing utility subject to a linear budget constraint p
x
x+p
y
y = I: There
could be any number of funky budget constraints in the presence of taxes, subsidies, rationing, coupons, etc.
The Lagrangian Method may, or may not, continue to nd the optimal bundle. If the portion of the budget
line the optimal bundle is on is unaected by a changing budget constraint, then clearly the optimal bundle
would not change.
Suppose there is a "kink" in the budget constraint which could be given by a tax applied to one good
if more than a certain amount is consumed of that good. Notice that the slope at the kink is undened as
the derivative does not exist at that point. Graphically, we see that the more the budget line tilts, the more
likely the optimal bundle is located at the kink. Intuitively, this is because the optimal bundle being at the
kink could be supported by a wide range of values for the MRS: If the slope of the upper portion of the
budget constraint is A and the slope of the lower portion is B; then an optimum at the kink could support
any MRS between A and B:
16
Figure 12: The optimal bundle being at the kink could be supported by a wide range of indierence curves
Figure 13: The optimal bundle is on a portion of the budget line that is unaected by the tax, demand does
not change in the presence of the tax.
17
Figure 14: Demand changes in the presence of a tax on y; more x is substituted for y:
Figure 15: A consumer with these preferences would demand the bundle at the kink point in the presence
of a tax
18
Consumption Curves
Once we have derived the demand for goods, there are a variety of curves that are interesting in the analysis
of economic problems.
Denition 8 The Income Consumption Curve (ICC) is the set of optimal bundles for all values of income
holding prices constant.
ICC = (x
1
(p; I); x
2
(p; I); :::; x
n
(p; I); :::))[I R given prices p R
1
Figure 16: An Income Consumption Curve (ICC) given preferences and the budget constraint
19
Figure 17: The ICC when good y is inferior at some income levels
Denition 9 The Engel Curve (EC) is the set of optimal demand for one good for all values of income
holding prices constant.
EC = (x
i
(p; I))[I R given prices p R
1
If the slope of the ICC is always positive, then both goods are normal, otherwise one is inferior. What
is the intuition for this result? Equivalently, if the slope of the EC is always positive, the good is normal.
Remark 10 The easiest way to derive the Engel curve is to solve for income as a function of the good.
20
Figure 18: The Engel Curve (EC) for good x
Figure 19: The EC for good y where y is inferior for some levels of income
21
Example 11 Perfect Substitutes u(x; y) = ax + by where
px
py
<
a
b
: Note that in this case x
=
I
px
so the
Engel curve for x is given by I = p
x
x which is a straight line from the origin with slope p
x
: y
= 0 so the
Engel curve for y is a straight line where y = 0: The ICC is the set (
I
px
; 0) for all values of income.
Figure 20: The ICC for perfect substitutes when
a
b
>
px
py
22
Figure 21: The EC for good x for perfect substitutes
Example 12 Perfect Complements u(x; y) = min(ax; by): Recall that the demand for good x is x
1
(p; I) =
I
px+py
a
b
so the Engel Curve for good x is given by I = (p
x
+p
y
a
b
)x; a straight line from the origin with slope
(p
x
+p
y
a
b
):
23
Figure 22: The ICC for perfect complements
Example 13 Quasi-linear u(x; y) = f(x) +y: Notice that the ICC and EC are a vertical line at some point
x: This is what makes quasi-linear preferences unique, good x is normal until a certain point and then there
is a zero income eect. An example could be any good where one becomes sated with the good, like DVD
rentals or pencils; at most income levels our consumption of the good is the same, only at very low income
levels would we begin to scale back.
24
Figure 23: The ICC for quasi-linear preferences
Figure 24: The EC for good x with quasi-linear preferences. There is no income eect for x at most incomes
25
Denition 14 The Price Consumption Curve (PCC) for good x
i
is the set of optimal bundles for all values
of p
i
holding income and all other prices constant.
PCC
xi
= (x
1
(p
i
; p
i
; I); x
2
(p
i
; p
i
; I); :::; x
n
(p
i
; p
i
; I); :::))[p
i
R given income I R and all other prices p
i
R
1
Why does the slope of this curve tell us whether the goods are substitutes or complements?
Figure 25: The Price Consumption Curve (PCC)
26
Figure 26: The Inverse Demand Curve for good x
Example 15 Draw the PCC for perfect substitutes when
px
py
<
a
b
Note that in the 2-good case, the PCC will begin at either
I
px
or
I
py
: This is because at a price of p
x
= ,
all income will be spent on good y: This also requires that good x and y are substitutes over some range
of prices (this range could be just the point where p
x
= as in the case of perfect complements) because
if all income is spent on y; and then at some point some income is spent on x; expenditure on y must go
down. This is unique to the 2-good case. A more general denition of substitutes and complements that we
will lean later will correct for this accounting fact, the denition of substitutes and complements that we are
currently using is technically what we call gross complements and gross substitutes.
Problem 16 Graph the demand function for good x when u(x; y) = ax +by
27
Figure 27: The PCC for perfect substitutes when
a
b
>
px
py
Homotheticity
Denition 17 Preferences are homothetic if whenever A % B; then A % B for > 0:
If preferences are homothetic if whenever you prefer one bundle to another, you always prefer twice
that bundle to twice the other bundle, or one-half that bundle to one-half the other bundle. Examples of
homothetic preferences include Cobb-Douglas, perfect substitutes, and perfect complements. If preferences
are homothetic, then
1. The ICC and ECs are straight lines from the origin.
2. If the optimal bundle demanded is x with income I; then the optimal bundled demanded with income
tI is tx:
3. A ray from the origin will intersect all indierence curves at the same angle (all ICs look the same no
matter how far they are from the origin).
4. MRS of an indierence curve only depends on the ratio of the goods, and not on their absolute value.
The last condition gives us a test for homotheticity, if MRS only depends on the ratio of two goods,
then preferences are homothetic. For example, the MRS for Cobb-Douglas preferences is
a
b
y
x
which does
not change if the ratio of
x
y
remains constant, so Cobb-Douglas preferences are homothetic.
28
Figure 28: When preferences are homothetic, any ray from the origin intersects all ICs at the same angle
Revealed Preference
So far, we have looked at how knowledge about preference allows us to derive optimal choice functions given
prices and income. Is there a way to go in the other direction, derive preferences given optimal choices? In
the real world, we observe choices, not preferences. It would be helpful to know if choices are consistent with
utility maximization and to be able to recover preferences from choices. Revealed preference allows us to
infer preferences from choices and requires minimal assumptions about the underlying preferences.
Suppose that you observe the optimal choice (the optimal bundle demanded) given prices and income
for the consumer, denoted x
(p; I): What can we say about this bundle compared to the other bundles of
goods in the choice set? If preferences are complete, we know that x
% y for all other y in the budget set,
otherwise there would be some y in the budget set such that y ~ x
and then y would have been chosen
instead of x
. We say that x
is "revealed preferred" to the other bundles in the budget set because the
consumer could have picked any of the other bundles, but picked x
:
If preferences are monotone, then we know that expenditure on x is equal to income, otherwise there
would be some bundle of goods with more of both goods that was aordable under prices and income and
that would have yielded more utility then x
:
To really get the power out of revealed preference, we need to prove a couple of propositions.
Proposition 18 If preferences are complete and monotone, then for every bundle z in the budget set that
is not on the budget line (p z < I); there is some good y on the budget line that is strictly preferred to z:
Proof: Take any bundle z such that p z < I: You can see that by drawing a 45 degree line upwards from
29
z; the line will eventually intersect the budget line. This bundle where the 45 degree line intersects the budget
line will have more of both goods and by monotonicity will be strictly preferred to z:
Proposition 19 If preferences are monotone, complete, and transitive, then x
~ z for all bundles z that
are in the budget set but not on the budget line (p z < I).
Proof: We know that x
% y for all bundles y in the budget set since x
is the optimal bundle and this
includes all bundles on the budget line. Take any bundle z in the budget set such that expenditure on z is
less than income, then by the previous proposition, there is some some bundle y satisfying y ~ z: We thus
know that x
% y ~ z and by transitivity, x
~ z:
Figure 29: The optimal choice x
is strictly preferred to all bundles in the budget set that are not on the
budget line.
These propositions simply say that if preferences are complete, monotone, and transitive, then x
is
strictly preferred to all bundles that cost less than x
. If these propositions are violated, then we know
that either preferences are not monotone, complete, and transitive, or that the consumer is not maximizing
utility. If choices do not violate these propositions, we say they satisfy the weak axiom of revealed preference.
Denition 20 Choices satisfy the Weak Axiom of Revealed Preference (WARP) if given optimal choice x
under prices p and income I and optimal choice y
under prices p
0
and income I
0
;
p y
_ I == p
0
x _ I
p y
< I == p
0
x > I
30
The logic of WARP (called "WARP" and not "the WARP") is that if y
was in the budget set and not
on the budget line when x
was chosen, by proposition 19 we know that x
~ y
: Therefore, when y
was
chosen, if x
was in the budget set, then utility is not being maximized (or preferences are not monotone,
transitive, or complete). Another way to think about it is if choices violate WARP, then there is no way you
can draw a set of indierence curves that are consistent with the optimal choices.
Remark 21 Preferences are assumed to be consistent, if preferences can change between periods then the
consumer may be utility maximizing, but choices between periods might violate WARP.
Figure 30: Choices are consistent with WARP. y
is aordable when x
is chosen, thus x
must be unaord-
able when y
is chosen; clearly x
is not in budget set 2.
31
Figure 31: Choices also consistent with WARP.
Figure 32: Choices not consistent with WARP; y
is in budget set 1 (but not on budget line 1) when x
chosen, thus x
~ y
: However, x
is in budget set 2 when y
is chosen, so WARP is violated.
32
Example 22 Cletus spends all his money on NASCAR tickets and beer. you have the following data on his
choices:
Income p
NASCAR
p
beer
q
NASCAR
q
beer
Period 1 $9 $9 $3
1
2
3
2
Period 2 $12 $4 $6 2
2
3
Assume Cletuss preferences are strictly monotone, transitive, and complete, the same over the three
months, and that he has no way to save or borrow across periods. Use a revealed preference approach to
establish whether his choices are consistent with utility maximization.
To nd whether his choices are consistent with WARP, we check the expenditure on each of the bundles
in each period under both prices:
Expenditure Bundle 1 Bundle 2
Period 1 Prices $9 $20
Period 2 Prices $11 $12
The optimal bundles satisfy expenditure = income in each period. To check if WARP is violated, we need
to check the aordability of the bundles when they are not chosen; if they are aordable when they were not
chosen, then the preferred bundle must be unaordable when they are chosen. Because p
2
x
= $11 < I
2
=
$12 , to satisfy WARP we must have that p
1
y
= $20 > I
1
= $12. These choices satisfy WARP and are
thus consistent with utility maximization.
Figure 33: The choices in example 1 are consistent with WARP
33
Example 23 Suppose instead that the optimal choices are:
Income p
NASCAR
p
beer
q
NASCAR
q
beer
Period 1 $9 $9 $3
1
2
3
2
Period 2 $12 $4 $6
1
4
11
6
These choices violate warp because p
1
y
= $
31
4
< I
1
= $9, but p
2
x
= $11 < I
2
= $12 . Because y
was strictly aordable when x was chosen, we must have x
~ y
; but in period 2, y
was chosen when x
was aordable, so these choices violate utility maximization (or the preferences are not complete, monotone,
transitive, or constant between periods).
Figure 34: The choices in example 2 are not consistent with WARP
34
